Measurement of Temperament in Infancy

3-, 6-, 9-, and 12-month-old subjects Scales widi adequate psychometnc and conceptual proper- bes were ... 141 children from 85 New York famihes were.
817KB taille 30 téléchargements 341 vues
Measurement of Temperament in Infancy Mary Klevjord Rothbart University of Oregon RoTHBABT, MABY KLEVJOHD Measurement of Temperament m Infancy CHILD DEVELOPMENT,

1981, 52, 569-578 Development of a caretaker-report instrument of the assessment of infant temperament is descnbed, and longitudinal findmgs are reported Temperament dimensions were selected for mvestigation from the work of Thomas, Chess et al, Escalona, Diamond, and others Conceptual analysis of scale definitions was earned out to eliminate conceptual overlap of scales, and item analysis was performed for 463 Infant Behavior Quesbonnaires filled out for 3-, 6-, 9-, and 12-month-old subjects Scales widi adequate psychometnc and conceptual properbes were developed for the following dimensions activity level, soothabilih', fear, distress to limitabons, smiling and laughter, and duration of orienting In longitudinal analyses, acbvity level and smiling and laughter scales revealed stability from 3 through 12 months, durabon of onenting and soothabihty showed less general stability, and fear and distress to limitabons showed stability only beyond the age of 6 months

Temperament as a psychobiological concept has been used m connecbon with the study of mdividual differences m other animal species (Diamond 1957), research on behavior genebcs (Goldsmith & Gottesman 1981), and the study of mdividual differences among human infants (Thomas & Chess 1977) Since the concept of temperament provides an mtegrative approach to the study of the development of individual differences (Rothbart & Derryberry, m press—a) and may eventually allow us to bace the relabon between children's early charactenstics and their social and cognitive development, the task of developing adequate measures for temperament is an important one Although psychologists, pediatricians, and parents have been to some extent aware of temperamental differences among infants, only recently have researchers attempted to develop techniques for assessing temperamental characteristics of infants beyond the neonatal period (Bates, Freeland, & Lounsbury 1979, Carey 1970, Carey & McDevitt 1978, Persson-Blennow & McNeil 1979, Scarr & Salapatek 1970, Thomas, Chess, Birch, Hertzig, & Kom 1963, Torgersen & Knnglen 1978, Pederson, Anderson, & Cam, Note 1) In this paper, temperament will be defined as individual differences in reacbvity and self-regulation Temperament

IS assumed to have a consbtubonal basis, with "constibibonal" defined as the relabvely enduring biological makeup of the individual, infiuenced over time by the interacbon of heredity, life experience, and maturabon (Rothbart & Derryberry, m press-b) All the studies listed above have used parent report to assess temperamental characteristics of the child, either via parent mterview, Q sort, or quesbonnaire Parent reports have the advantage of ublizmg the parent's extensive observabons of the child over a Wide variety of home situabons in assessment of children's temperamental characteristics To the extent that parent report instruments overcome possible response biases (descnbed by Bates [m press]), they provide a characterization of the infant's typical temperament hehavior as seen in the home We have cauboned elsewhere, however, that parent report measures do not provide an assessment of infant temperament that is independent of the child's current environment, rather, they measure the infant's behavior as seen in the home (Rothbart & Derryberry, in press-a) Behavior in the home reflects the result of the mteracbon between patterns of social sbmulabon in the home and the child's own temperamental patterns of reacbvity and

This research was parbally supported by grant 5 ROl MH 26674-04 from the Nabonal Insbtute of Pubhc HealUi. I wish to thank the following persons for their contnbubons to the research Lita Furby, Susanna Kelly, and Joan Hamilton Portions of this study were previously reported at Ae meetings of the Society for Research m Child Development, New Orleans, March 1977 Requests for repnnts, copies of the Infant Behavior Quesbonnaire and sconng procedures should be sent to Mary K Rothbart, Department of Psychology, University of Oregon, Eugene, Oregon 97403 ICkiU Dtfthtmmt, 1981, 52, 569-578 © 1981 by the Soaety lor Research in Child Development, Inc

570

Child Development

self-regulabon Parent report measures of temperament may thus be seen as assessing temperament as it IS demonstrated withm the infant-caregiver system of interacbon Viewed m this way, parent-reported temperament allows us to make group and developmental comparisons, for example, examining developmental sex differences in temperament m the home or charactenzmg the behavior of Down's syndrome infants m comparison with matched normal infants (Hanson 1979), but we must reahze in making these comparisons that our temperament measure is not mdependent of the children's home environment The first attempt at large-scale assessment of infant temperament via parent report was the work of Thomas, Chess, and their colleagues (Thomas & Chess 1977, Thomas, Chess, & Birch 1968, Thomas et al 1963) Parents of 141 children from 85 New York famihes were mterviewed durmg the first 2 years of their infants' lives After a content analysis of their first 22 interviews, Thomas et al (1963) chose the following temperament charactensbcs for further analysis activity level, rhythmicity, approach, adaptabihty, threshold, intensity, attenbon span, distracbbihty, and persistence These dimensions were chosen because they were found to be scorable for all protocols and because they allowed a sufficiently wide distnbution of scores for furtber study Two-hour interviews were scored on three-point scales for all behavior items corresponding to given temperament vanables Preponderance scores were then determined by tallying frequencies of high, medium, and low responses for each child The Thomas et al studies have been highly mfluenbal m the study of infant temperament, although there are some problems in interpretmg their results First, it is impossible to determine the extent of homogeneity within any given behavior scale for example, if acbvity is menboned rarely by the mother, a high score on activity could result from activity only dunng the feeding situabon Temperament scales are also constrained by the three-point scale used m protocol sconng Second, the subjects ranged m age from 2 months to 6 months at the bme of the first mterview, with subsequent interviews beld at 3-month periods Some of the subjects were thus considerably older than other subjects dunng all phases of the study, and differences related to age may have confounded findings of actual individual differences among the children Finally, the New York sample was highly re-

stricted with respect to SES and ethnic group, and 47* of the families contributed more than one subject to the project Since the Thomas et al studies, there have been several attempts to devise parent report instruments based on the New York temperament interviews Scarr and Salapatek (1970) chose Items based on examples from ITiomas et al's 1963 book, to which mothers were asked to respond "mostly true" or "mostly false" Children m the study ranged m age from 2 months to 24 months Item analyses were carried out for 70 mothers' responses to these items, but the investigators reported that internal consistency of the scales "was found to be rather low " The Scarr-Salapatek questionnaire has not been used in subsequent studies of infant temperament A quesbonnaire devised by Carey (1970), however, has been used m the Rochester infancy study (Bakow, Sameroff, Kelly, & 2;ax, Note 2) and extensively m other temperament research (reviewed by 'Thomas & Chess [1977]) Carey originally developed a 70-item quesbonnaire pnmanly as a clmical screening device for pediatncians The quesbons were also based on the Thomas et al work, and nine scales measured the charactensbcs defined in the New York study No item analysis was earned out for Carey's first instrument More recently, however, Carey and McDevitt (1978) have revised the nine scales, usmg 203 4-8-monthold infants for a standardizabon populabon Internal consistency of this revised scale ranges from 49 to 71 (statistic unspecified), with a median of 57 Pederson, Anderson, and Cam (Note 1) have developed a Q sort measure for assessing parents' percepbons of Thomas et al's (1963, 1968) nine categories of temperament They report corrected spht-half reliabilibes for the dimensions of acbvity, rhythmicity, adaptability, approach, and positive mood ranging from 54 to 69, wath a median of 60 They had less success measuring the dimensions of threshold, persistence, distracbbihty, and intensity, with corrected split-half correlations ranging from 31 to 48, with a median of 40 A quesbonnaire in Swedish (Persson-Blennow & McNeil 1979) and an mterview schedule in Norwegian (Torgersen & Knnglen 1978) based on the Thomas et al (1963, 1968) categories have also been developed Bates et al (1979) have developed a caretaker-rabng instrument, the Infant Characteristics Quesbonnaire, consisbng of 24 items

Mary Klevjord Rothbart scored on seven-point scales Responses of 322 mothers of 4-6-month-oId infants were factor analyzed, yielding four factors labeled by Bates et al as (1) fussy/difficult, (2) unadaptable, (3) dull, and (4) unpredictable Internal consistency estimates for the factor scales with a cross-vahdabonal sample of 2V = 196 ranged from 39 to 79, with a median coefficient « of 63

Infant Temperament Dimensions The purpose of the present project in developing a parent report mstrument of infant temperament was to develop a psychometrically adequate instrument as reflected by high internal rehabihty that would measure not only the Thomas, Chess et al (1963, 1968) dimensions, but would tap otber aspects of reacbvity and self-regulation that had been identified as involving individual differences with a possible constitubonal basis In addition, we wished to ldenbfy dimensions of temperament that were conceptually independent, that is, involving no overlap among operational definitions Both the Thomas et al work and the work by Carey (1970, Carey & McDevitt 1978) had involved such overlap, and :t was hoped that, by developing conceptually disbnct temperament dimensions, we could adequately explore correlations between dimensions without lnflabng them by using similar items on scales with different names It was also hoped that we could study both developmental continuity and change m children's patterns of reacbvity and self-regulation as observed over time in tbe home Choice of lnibal temperament dimensions for the present research was therefore based upon the work of Thomas et al (1963, 1968) but was influenced by other research and theory as well, notably the work of Escalona (1968) and Shirley (1933), sbidies of temperament and behavior genetics m animals and humans (reviewed by Diamond [1957]), and Iongitudmal studies of personality in human subjects On the basis of this review, 11 temperament dimensions were selected for lnibal investigation Four of the variables may be seen as tapping general charactenstics of response as assessed across differing sensory receptors and response channels, these include threshold, mtensity, and adaptabihty (soothabihty) of response, and rhythmiaty Seven vanables involve activabon of more restncted responses acbvity level, fear, distress to limitations, overall negabve emobonality, smilmg

571

and laughter, duration of onenbng, and distracbbihty The short literature reviews given with each variable below are not inclusive, studies are cited to point up the importance of the dimension to temperament research, more inclusive reviews may be found m Rothbart and Derryberry (in press-a) Actwtty level—Level of gross motor activity has probably been the most widely studied temperamental characteristic m young children ( e g , Escalona 1968, Fries & Woolf 1953, Richards & Newberry 1938, Schaffer 1966, and Thomas et al 1963), and animal behavior genetics work (Fuller & Thompson 1978) as well as human twm study research have suggested that acbvity level may be hentable (Matheny, Dolan, & Wilson 1976, Scarr 1966, WiUerman 1973, Willerman & Plomin 1973) Buss and Plomin (1975) reviewed longitudinal studies on acbvity level, concluding that activity level shows stability as a personality dimension only after the period of infancy In the early months, however, Birns, Barten, and Bridger (1969) bave found consistency of activity between infants of 1 and 4 montbs Fear (distress and extended latency to approach tntense or novel stimuli) —^An approach scale (defined as acceptance or rejection of new objects or persons) was mcluded in the Thomas et al study (1963), and reacbOn to intense or novel stimuli appears to be a highly promising dimension for the study of temperament m children Infants do show fear responses (Bronson 1968, Scarr & Salapatek 1970), and animal studies have indicated the existence of inherited mdividual differences in fearfulness in rats, dogs, and chimpanzees (Fuller & Thompson 1978, Hall 1951, Yerkes & Yerkes 1936) High heritability coefficients have also been found on anxiety rabngs for both child and adult twin pairs (Gottesman 1963, Scarr 1966) Distress to limttatums—This variable has been chiefly studied in connecbon with persistence and goal onentation, since individual differences in mfants' reacbons to fnistrabng condibons may be easily observed Kramer and Rosenblum (1970), for example, studied the reacbons of 1-year-olds to the placing of a glass barrier between the child and an attractive toy Three kmds of response were observed some infants were persistent and managed to reach the toy, some shifted their attention to another environmental focus without reaching the toy, a third group became distressed and quickly

572

Child Development

lost interest m the toy Evidence for heritabihty of persistence of goal onentation has been put forward by Torgersen and Krmglen (1978) and Goldsmith and Gottesman (1981), and Wilson, Brown, and Matheny (Iff71) have reported that "temper frequency" formed the core of a cluster of distress items in child behavior as reported by maternal interviews from 6 to 36 months Aggressiveness is a bait for which animal crossbreeding has been successful (Hall & Klein 1942), but we do not know the precise relabon between an aggressive response and a distress response to frustration In addibon to the distress scales listed above, an attempt was made to assess a general dimension of negative emotionahty Threshold, tntenstty, and adaptabdity (soothabtlity) of response —^The Thomas et al (1963) study included scales for intensity, threshold, adaptabihty, and mood A study by Freedman (1971) usmg the Brazelton assessment of neonatal temperament compared 24 Chmese-Amencan newboms with 24 EuropeanAmencan newhoms, finding that the EuropeanAmencan infants reached a peak of excitement earher in the assessment penod, wavenng between contentment and upset, while the Chinese-Amencans scored on the calmer and steadier side of these items, were more easily consoled, and were more hkely to soothe themselves Bims et al (1969) found stability on ratings of imtabihty, sensitivity, and soothabihty over the first 4 months In the Thomas et al (1968) study, mtensity of mdividual response showed relabvely high stabihty over time and showed consistent posibve correlabons with other temperament scales However, many of these correlations may be artifactual, since the same mother's statement could be coded for any number of scales For example, mtensity of response, threshold, and mood were often scored from the same statements Smtltng and laughter —Smihng and laughter have been identified as indicators of arousal under safe condibons (Rothbart 1973, Sroufe & Waters 1976) and deserve more thorough explorabon as lndicabons of emotional reacbvity than has been the case Buss and Plomm (1975), for example, have argued that posibve emobons do not reflect consbtubonally based emobonahty Freedman (1971), however, presents evidence that pairs of identical twins are more ahke than pairs of fraternal tvuns in smihng dunng the first year Wilson, Brown, and Matheny (1971) also report that smihng consbtutes part of a sociabihty cluster

extracted from maternal reports of child behavior from 6 months through 72 months of age Rhythmtctty —Rhythmicity of sleep and hunger cycles and bowel movements was measured by Tbomas et al and described as a characteristic of the "easy child " Duratum of orientmg and dtstractibiltty — Thomas et al (1963, 1968) measured both attenbon span and distractibihty m infants and young children, thereby pombng up the possibihty of important individual differences m the development of attention Attentional vanables are of interest m that several researchers using different age groups and quite different procedures (Cohen 1975, McCall & Kagan 1970, McCall & Melson 1969, Paden 1975, Self 1975) have rejwrted that mfants differ in their rate of habituabon and, hence, durabon of onentmg over tnals Wilson, Brown, and Matheny (1971) have reported attenbon span to be negatively related to a cluster including temper frequency and intensity from 6 through 18 months of age In the present study, preparation of the scales involved both conceptual and item analysis After tbese analyses, six scales were developed with adequate conceptual and psychometric characteristics activity level, smiling and laughter (ongmally labeled posibve emobonahty), fear, distress to limitabons (onginally labeled anger/frusbabon), soothabih^, and duration of orienting (originally labeled persistence)

Method Initial Items for the Infant Behavior Questionnaire were developed from tbe work of Shirley (1933), Thomas et al (1963), and others, as well as from mtensive interviews with 26 parents of 3-, 6-, 9-, and 12-month-old infants Items were selected to assess one of 11 different dimensions of temperament listed above In order to avoid the real problems often associated with parent report (Weick 1968, Yarrow 1963), we wished to avoid asking parents either to make global judgments of their child's behavior or to attempt to recollect occasions of child behavior from the distant past We did not wish to ask parents to make comparabve judgments about their mfants—^for example, "my child is extremely active"— since lnformabon about the behavior of other infants might not be available to them Items m the instrument were therefore carefully worded to refer to specific behaviors, and care-

Mary Klevjord Roihbart

573

takers were asked to respond on *he basis of the infant's specific behavior dunr'-? the past week Data from the experimental literature (Hasher & Zacks 1979) suggest that (a) frequency judgments of even nonattended iiems tend to be remembered witb some accuracy, (b) frequency judgments tend to be made independently from recency or durabon mformabon, and (c) giving explicit instructions for attending and frequency of occurrence of items does not greatly improve performance m memory for frequency An example of the questionnaire items used m the item analysis is given below The responses were coded on a scale from 1 to 6 1 = never, 2 = onee or twice, 3 = less than half the bme, 4 = about half the tune, 5 = more than half the time, 6 =:: almost always, X = does not apply

a new food or Lquid or left with a babysitter Although it has not been possible to assess response bias for a given subject (smce, e g , the tendency to use only the high end of the scale may be an altogether valid indicator of a particular child's bebavior), our empbasis on caretakers' estimabng the relabve frequency of concrete behaviors dunng the past week tends to minimize the possibihty of bias In addition, the inclusion of a number of reverse items minimizes effects of response bias, with scores for these Items reversed when compubng the relevant scale score For example, a reverse item for the acbvity scale is "Dunng sleep, how often did your bahy sleep m one position only!*" A score of 6 on this item reflects low acbvity, and is translated into a 1 when activity scores are computed

Dunng the past week, when often did your baby Wave his/her arms and 1 2 3 4 Cry? 1 2 3 4 Smile or laugh? 1 2 3 4

Two kmds of analyses have been performed on these scales conceptual and item analyses The conceptual analysis has involved (a) development of precise operabonal definitions of each dimension of temperament and (b) subsequent ehmmabon of scales that conceptually overlap with each other Two of the ongmal 11 scales were discarded because of unavoidable conceptual overlap with other temperament scales negative emobonality showed overlap with both the distress to limitations and fear scales, and we were not able to formulate a mutually exclusive operational definibon of distracbbilitv Definitions of temperament dimensions remammg after complebon of conceptual and Item analyses are listed in table 1 Several of the temperament dimensions have also been relabeled in an attempt to reflect only information conveyed m the operabonal definitions and avoid insofar as possible value judgments inherent in scale names, for example, names like "persistence" and "posibve emotionality "

bemg undressed, how kick? 5 6

X

5

6

X

5

6

X

We have recently modified the response options, making them more symmetncal and more nearly an interval scale The improved response opbons are as follows 1 — never, 2 — very rarely, 3 = less than half the bme, 4 = about half tbe bme, 5 — more than half the bme, 6 = almost always, 7 = always, X = does not apply A response opbon for mdicatmg that an item "does not apply" was included because some of the specific situabons descnbed in the quesbonnaire may not have occurred during the previous week for a parbcular infant, for example, the infant may not have been given

TABLE 1 TEMPEKAMENT DIMENSION DEFINITIONS

Dimension Activity level Smiling and laughter Fear Distress to limitations Sootkabihty Duration of onenUng

Definition Child's gross motor activity, mdudmg movement of arms and legs, squirming, and locomotor activity Smiling or laughter from the child m any situation The child's distress and/or extended latency to approach an intense or novel stimulus Child's fussmg, crymg, or showing distress while (a) waiting for food, (6) refusing a food, (